Research
Original Investigation
Comparison of Application of the ACC/AHA Guidelines, Adult Treatment Panel III Guidelines, and European Society of Cardiology Guidelines for Cardiovascular Disease Prevention in a European Cohort Maryam Kavousi, MD, PhD; Maarten J. G. Leening, MD, MSc; David Nanchen, MD, MSc; Philip Greenland, MD; Ian M. Graham, MD; Ewout W. Steyerberg, PhD; M. Arfan Ikram, MD, PhD; Bruno H. Stricker, MMed, PhD; Albert Hofman, MD, PhD; Oscar H. Franco, MD, PhD Editorial page 1403 IMPORTANCE The 2013 American College of Cardiology/American Heart Association
(ACC/AHA) guidelines introduced a prediction model and lowered the threshold for treatment with statins to a 7.5% 10-year hard atherosclerotic cardiovascular disease (ASCVD) risk. Implications of the new guideline’s threshold and model have not been addressed in non-US populations or compared with previous guidelines. OBJECTIVE To determine population-wide implications of the ACC/AHA, the Adult Treatment Panel III (ATP-III), and the European Society of Cardiology (ESC) guidelines using a cohort of Dutch individuals aged 55 years or older.
Related articles pages 1406 and 1424 Supplemental content at jama.com CME Quiz at jamanetworkcme.com and CME Questions page 1442
DESIGN, SETTING, AND PARTICIPANTS We included 4854 Rotterdam Study participants recruited in 1997-2001. We calculated 10-year risks for “hard” ASCVD events (including fatal and nonfatal coronary heart disease [CHD] and stroke) (ACC/AHA), hard CHD events (fatal and nonfatal myocardial infarction, CHD mortality) (ATP-III), and atherosclerotic CVD mortality (ESC). MAIN OUTCOMES AND MEASURES Events were assessed until January 1, 2012. Per guideline, we calculated proportions of individuals for whom statins would be recommended and determined calibration and discrimination of risk models. RESULTS The mean age was 65.5 (SD, 5.2) years. Statins would be recommended for 96.4% (95% CI, 95.4%-97.1%; n = 1825) of men and 65.8% (95% CI, 63.8%-67.7%; n = 1523) of women by the ACC/AHA, 52.0% (95% CI, 49.8%-54.3%; n = 985) of men and 35.5% (95% CI, 33.5%-37.5%; n = 821) of women by the ATP-III, and 66.1% (95% CI, 64.0%-68.3%; n = 1253) of men and 39.1% (95% CI, 37.1%-41.2%; n = 906) of women by ESC guidelines. With the ACC/AHA model, average predicted risk vs observed cumulative incidence of hard ASCVD events was 21.5% (95% CI, 20.9%-22.1%) vs 12.7% (95% CI, 11.1%-14.5%) for men (192 events) and 11.6% (95% CI, 11.2%-12.0%) vs 7.9% (95% CI, 6.7%-9.2%) for women (151 events). Similar overestimation occurred with the ATP-III model (98 events in men and 62 events in women) and ESC model (50 events in men and 37 events in women). The C statistic was 0.67 (95% CI, 0.63-0.71) in men and 0.68 (95% CI, 0.64-0.73) in women for hard ASCVD (ACC/AHA), 0.67 (95% CI, 0.62-0.72) in men and 0.69 (95% CI, 0.63-0.75) in women for hard CHD (ATP-III), and 0.76 (95% CI, 0.70-0.82) in men and 0.77 (95% CI, 0.71-0.83) in women for CVD mortality (ESC). CONCLUSIONS AND RELEVANCE In this European population aged 55 years or older, proportions of individuals eligible for statins differed substantially among the guidelines. The ACC/AHA guideline would recommend statins for nearly all men and two-thirds of women, proportions exceeding those with the ATP-III or ESC guidelines. All 3 risk models provided poor calibration and moderate to good discrimination. Improving risk predictions and setting appropriate population-wide thresholds are necessary to facilitate better clinical decision making.
JAMA. 2014;311(14):1416-1423. doi:10.1001/jama.2014.2632 Published online March 29, 2014. 1416
Author Affiliations: Author affiliations are listed at the end of this article. Corresponding Author: Maryam Kavousi, MD, PhD, Department of Epidemiology, Erasmus University Medical Center, Rotterdam, PO Box 2040, 3000 CA Rotterdam, the Netherlands (m.kavousi @erasmusmc.nl). jama.com
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
Comparison of Guidelines for CVD Prevention
P
revention of cardiovascular disease (CVD), the leading cause of death worldwide,1,2 remains feasible3 yet suboptimal. The common approach in CVD primary prevention is to identify individuals at high enough risk for cardiovascular events to justify targeting them for more intensive lifestyle interventions, pharmacological interventions, or both. The CVD prevention guidelines developed by the National Cholesterol Education Program expert panel,4 succeeded by the American College of Cardiology/American Heart Association (ACC/AHA) task force5, and the European Society of Cardiology (ESC)6 are the major guidelines influencing clinical practice. While the Adult Treatment Panel III (ATP-III) guidelines were based on the 10-year risk of coronary heart disease (CHD) only,4 the ACC/AHA guidelines broaden to comprise risk of all hard atherosclerotic CVD (ASCVD), including CHD and stroke,5 using the Pooled Cohort equations.7 An additional substantial change in the US guideline is a lower risk threshold for statin treatment in asymptomatic individuals from 20% CHD risk in the ATP-III guidelines4 to 7.5% ASCVD risk in the new guidelines.5 The potential implications of the ACC/AHA guidelines in largely widening the populations endorsed for treatment and the accuracy of the ACC/AHA risk calculator have received much attention.8-12 To be clinically useful, risk prediction models should provide good discrimination. Because decisions for statin treatment are based on an individual’s absolute risk, calibration of the risk prediction models as well as the risk threshold for treatment are important. Varying approaches to CVD risk estimation and application of different criteria for therapeutic recommendations would translate into substantial differences in proportions of individuals qualifying for treatment at a population level. We therefore aimed to determine implications of the ACC/AHA, the ATP-III, and the ESC guidelines in a prospective cohort of Dutch individuals aged 55 years or older. Our first aim was to determine what proportion of the population would be treated based on each guideline. We then sought to examine discrimination and calibration of the 3 risk prediction models underlying these guidelines.
Methods Study Population Analyses were performed within the framework of the Rotterdam Study, a prospective population-based cohort study among persons aged 55 years or older in the Ommoord district of Rotterdam, the Netherlands. The rationale and design of the Rotterdam Study have been described elsewhere.13 The baseline examination took place in 1990-1993 (RS-I). In 2000, the cohort was extended to include inhabitants who reached the age of 55 years in 1990-2000 and persons aged 55 years or older who migrated into the research area (RS-II). The Rotterdam Study was approved by the Medical Ethics Committee of the Erasmus Medical Center and all participants provided written informed consent. The present study used data from the third examination of the original cohort (RS-I, recruited 1997-1999) and the first examination of the extended cohort (RS-II, recruited 2000-
Original Investigation Research
2001). Among the participants aged 75 years or younger, there were 2209 men and 2645 women with measurements required for the analyses. Among these individuals, 315 men and 330 women were receiving statin treatment at baseline and therefore were excluded from the population for whom the eligibility for treatment based on each guideline was assessed. For further analyses on examining the performance of each risk scoring model, exclusions were made using the criteria from each guideline.
Main Outcome Measures and Follow-up Main outcomes were hard ASCVD, composed of fatal and nonfatal myocardial infarction (MI), other CHD mortality, and stroke; hard CHD, composed of fatal and nonfatal MI and CHD mortality; and atherosclerotic CVD mortality.14,15 Prevalent CVD was defined as a history of MI, coronary or other arterial revascularization, stroke or focal transient ischemic attack, or heart failure. Events were assessed until January 1, 2012. A complete description of the methods for measurement of cardiovascular risk factors, definitions of the outcomes, and details regarding the follow-up time is provided in the eAppendix in the Supplement.
Statistical Analyses We calculated the 10-year risk of hard ASCVD events for each individual based on age, systolic blood pressure, treatment of hypertension, total and high-density lipoprotein (HDL) cholesterol levels, current smoking, and history of diabetes mellitus, using the sex-specific parameters from the ACC/AHA Pooled Cohort equations.7 We used the recommended 5% and 7.5% risk thresholds for categorization of the 2 respective categories of “treatment considered” and “treatment recommended.”5 To comply with the ACC/AHA guideline,5 the risk estimation for hard ASCVD was calculated among individuals who were not receiving lipid-lowering medication, were free of CVD at baseline, and had low-density lipoprotein (LDL) cholesterol levels below 190 mg/dL. Using the continuous ATP-III risk prediction model based on age, systolic blood pressure, treatment of hypertension, total and HDL cholesterol levels, and current smoking,16 we also calculated the 10-year risk of hard CHD for the individuals who were not receiving lipid-lowering medication and were free of CVD and diabetes mellitus, to comply with the ATP-III guideline.4 The risk thresholds used for categorization were 10% and 20%, corresponding to the cutoff points for defining the intermediate- and high-risk categories by the ATP-III guideline.4 The 10-year risk of CVD mortality for each participant was based on age, systolic blood pressure, total cholesterol levels, and current smoking using the sex-specific intercepts and regression coefficients from the SCORE equation for low-risk European countries.17 We used the recommended 1%, 5%, and 10% risk thresholds, corresponding to the cutoff points for defining the moderate-risk, high-risk, and very-high-risk groups, respectively, based on the ESC guideline.6,18 To comply with the ESC guideline, the SCORE risk estimation was performed among the individuals who were not receiving lipid-lowering medication at baseline and were free of CVD, diabetes melli-
jama.com
JAMA April 9, 2014 Volume 311, Number 14
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
1417
Research Original Investigation
Comparison of Guidelines for CVD Prevention
Figure 1. Inclusion/Exclusion Criteria for Rotterdam Study Participants for Assessment of Different Guideline Recommendations and Risk Prediction Models ACC/AHA Guideline5
ATP-III Guideline4
4854 Total population 2209 Men 2645 Women
4854 Total population 2209 Men 2645 Women
645 Excluded for statin use at baseline 315 Men 330 Women 4209 Included in treatment recommendations of ACC/AHA guideline 1894 Men 2315 Women
4854 Total population 2209 Men 2645 Women
645 Excluded for statin use at baseline 315 Men 330 Women 4209 Included in treatment recommendations of ATP-III 1894 Men 2315 Women
776 Excluded for prevalent CVD or LDL-C >190 mg/dL (>4.9 mmol/L) 381 Men 395 Women
645 Excluded for statin use at baseline 315 Men 330 Women 4209 Included in treatment recommendations of ESC guideline 1894 Men 2315 Women
802 Excluded for prevalent CVD or DM 463 Men 339 Women
1027 Excluded for prevalent CVD, DM, or CKD 528 Men 499 Women
3433 Included in risk estimation for hard ASCVDa 1513 Men 1920 Women
3407 Included in risk estimation for hard CHDb 1431 Men 1976 Women
3182 Included in risk estimation for CVD mortality 1366 Men 1816 Women
343 Developed hard ASCVD during 10-y follow-up 192 Men 86 Stroke 72 Nonfatal MI 26 Fatal CHD 8 Fatal MI 151 Women 92 Stroke 41 Nonfatal MI 15 Fatal CHD 3 Fatal MI
160 Developed hard CHD during 10-y follow-up 98 Men 65 Nonfatal MI 22 Fatal CHD 11 Fatal MI 62 Women 43 Nonfatal MI 16 Fatal CHD 3 Fatal MI
87 CVD deaths during 10-y follow-up 50 Men 37 Women
ACC/AHA indicates American College of Cardiology/American Heart Association; ASCVD, atherosclerotic cardiovascular disease; ATP-III, Adult Treatment Panel III; CHD, coronary heart disease; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; ESC, European Society of
tus, and chronic kidney disease (CKD).6 Figure 1 describes the inclusion and exclusion criteria for different risk prediction models. Based on each guideline, we formed 3 categories of treatment: “treatment recommended,” “treatment considered,” and “no treatment.” eTables 1 through 3 in the Supplement describe the criteria used to form these 3 treatment categories by each guideline. We assessed the discrimination and calibration of each risk prediction model in our population. Discrimination refers to ability of the model to assign a higher risk to individuals who develop the outcome of interest compared with those who remain free of disease. The discriminative performance of each risk-scoring model was assessed using the C statistic. Calibration is the agreement between the predicted probabilities of disease, based on the risk prediction model, and the actual incidence of events in the population. To assess the calibration of each risk prediction model, the average predicted 10-year risks for each risk function were compared with the average 10-year observed risks (ie, cumulative incidence of the event). Calibration plots were generated to assess the agreement between the predicted and observed risks over the entire range. 1418
ESC Guideline6
Cardiology; LDL-C, low-density lipoprotein cholesterol; and MI, myocardial infarction. a Hard ASCVD includes fatal CHD, nonfatal CHD, and stroke. b Hard CHD includes fatal myocardial infarction, nonfatal MI, and CHD mortality.
Results Baseline characteristics of the participants are presented in Table 1. The mean age of the participants was 65.5 (SD, 5.2) years and 54.5% were women. Based on the ACC/AHA guideline,5 the “treatment recommended” group included 96.4% (95% CI, 95.4%-97.1%; n = 1825) of men and 65.8% (95% CI, 63.8%-67.7%; n = 1523) of women while the “treatment considered” group included 3.3% (95% CI, 2.6%-4.2%; n = 63) of men and 14.2% (95% CI, 12.8%-15.7%; n = 330) of women. Only 0.3% of men (95% CI, 0.1%-0.7%; n = 6) and 20.0% (95% CI, 18.3%-21.6%; n = 462) of women were categorized in the “no treatment” group (Table 2 and eTable 1 in the Supplement). Using the ATP-III guideline, 4 52.0% (95% CI, 49.8%54.3%; n = 985) of men and 35.5% (95% CI, 33.5%-37.5%; n = 821) of women were categorized in the “treatment recommended” group, while the “treatment considered” group included 14.2% (95% CI, 12.6%-15.8%; n = 269) of men and 14.1% (95% CI, 12.7%-15.6%; n = 326) of women. The “no treatment” category included the remaining 33.8% (95% CI,
JAMA April 9, 2014 Volume 311, Number 14
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
jama.com
Comparison of Guidelines for CVD Prevention
Original Investigation Research
31.7%-35.9%; n = 640) of men and 50.4% (95% CI, 48.4%52.5%; n = 1168) of women (Table 2 and eTable 2 in the Supplement). Based on the ESC guideline, 6 66.1% (95% CI, 64.0%68.3%; n = 1253) of men and 39.1% (95% CI, 37.1%-41.2%; n = 906) of women were included in the “treatment recommended” category. The “treatment considered” group comprised 31.6% (95% CI, 29.5%-33.7%; n = 598) of men and 51.4% (95% CI, 49.3%-53.4%; n = 1189) of women. Only 2.3% (95% CI, 1.6%-2.9%; n = 43) of men and 9.5% (95% CI, 8.3%-10.8%; n = 220) of women were assigned to the “no treatment” category (Table 2 and eTable 3 in the Supplement). eFigure 1 in the Supplement presents the treatment recommendations based on the 3 guidelines for the populations younger than 65 years and aged 65 years or older. The data suggest that almost all men older than 55 years and nearly all women older than 65 years are recommended for statin treatment based on the new ACC/AHA guideline. eTables 1 through 3 in the Supplement show that while all men and women with prevalent CVD were categorized in the “treatment recommended” group by the ACC/AHA guideline (eTable 1 in the Supplement), 12.9% of men and 4.2% of women with clinical CHD and CHD risk equivalents were categorized in the “treatment considered” or “no treatment” category based on the ATP-III guideline (eTable 2 in the Supplement). Using the ESC guideline, a small group of individuals with clinical CVD and its risk equivalents (0.6% of men and 0.4% of women) were categorized in the “treatment considered” group (eTable 3 in the Supplement). eTables 4 through 6 in the Supplement provide the description of the proportion of the population to whom each risk estimation model was applied. Among 1513 men and 1920 women included for ASCVD risk prediction (ACC/AHA), 192 men and 151 women developed hard ASCVD over 10-year follow-up. Among 1431 men and 1976 women included for CHD risk prediction (ATP-III), hard CHD occurred in 98 men and 62 women over 10-year follow-up. Among 1366 men and 1816 women included for CVD mortality risk prediction (ESC), 50 men and 37 women died of atherosclerotic CVD over 10-year follow-up. For all outcomes studied, follow-up time was truncated at 10 years for individuals with a longer follow-up time than 10 years. After calculating the 10-year risk for individuals based on each risk prediction model, we first assessed the discriminative ability of each model. The C statistic for the ACC/AHA
model was 0.67 (95% CI, 0.63-0.71) for men and 0.68 (95% CI, 0.64-0.73) for women for hard ASCVD. Use of the ATP-III risk prediction model resulted in a C statistic of 0.67 (95% CI, 0.620.72) for men and 0.69 (95% CI, 0.63-0.75) for women for hard CHD. Using the SCORE equation (ESC), the C statistic was 0.76 (95% CI, 0.70-0.82) for men and 0.77 (95% CI, 0.71-0.83) for women for CVD mortality. We then assessed the calibration of each risk prediction model. Figure 2 compares the average 10-year risks predicted by the ACC/AHA, ATP-III, or SCORE (ESC) risk prediction models with the observed 10-year risks (ie, cumulative incidence of events) in each risk category. Calibration was poor for all 3 models; the ACC/AHA (Figure 2A), the ATP-III (Figure 2B), and the SCORE equation (Figure 2C) overestimated the 10-year risk among men and women across all risk categories. eTable 7 in the Supplement details the percentage of population at different categories of risk using each risk prediction model. The average predicted risks vs observed cumulative incidence of
Table 1. Characteristics of the Study Population at Baseline Men (n = 2209)
Characteristics Age, mean (SD), y
65.5 (5.3)
65.4 (5.2)
Systolic
143 (21)
140 (21)
Diastolic
79 (11)
76 (11)
468 (21.2)
643 (24.3)
Blood pressure, mean (SD), mm Hg
Antihypertensive treatment, No. (%) Body mass index, mean (SD)a
26.7 (3.3)
27.3 (4.5)
Total cholesterol, mean (SD), mg/dL [mmol/L]
216.2 (37.1) [5.60 (0.96)]
232.7 (35.7) [6.03 (0.92)]
HDL cholesterol, mean (SD), mg/dL [mmol/L]
47.7 (12.1) [1.24 (0.31)]
58.1 (14.9) [1.50 (0.39)]
LDL cholesterol, mean (SD), mg/dL [mmol/L]
140.2 (34.4) [3.63 (0.89)]
147.9 (34.4) [3.83 (0.89)]
Statin treatment at baseline, No. (%)b
315 (14.3)
330 (12.5)
Current smoking, No. (%)
437 (19.8)
522 (19.7)
Diabetes mellitus, No. (%)
315 (14.3)
282 (10.7)
Chronic kidney disease, No. (%)
139 (6.3)
226 (8.5)
Prevalent CVD, No. (%)
414 (18.7)
186 (7.0)
Abbreviations: CVD, cardiovascular disease; HDL, high-density lipoprotein; LDL, low-density lipoprotein. a
Calculated as weight in kilograms divided by height in meters squared
b
No. (%) of men and women receiving lipid-lowering medication at baseline. (Statins constituted 96% of all lipid-lowering medications at baseline).
Table 2. Treatment Recommendations Based on Different Guidelines Guidelinea Treatment Categories
ACC/AHA
5
ATP-III4
ESC6
52.0 (49.8-54.3)
66.1 (64.0-68.3)
b
Men (n = 1894)
Treatment recommended
96.4 (95.4-97.1)
Women (n = 2645)
Treatment considered
3.3 (2.6-4.2)
14.2 (12.6-15.8)
31.6 (29.5-33.7)
No treatment
0.3 (0.1-0.7)
33.8 (31.7-35.9)
2.3 (1.6-2.9)
Treatment recommended
65.8 (63.8-67.7)
35.5 (33.5-37.5)
39.1 (37.1-41.2)
Treatment considered
14.2 (12.8-15.7)
14.1 (12.7-15.6)
51.4 (49.3-53.4)
No treatment
20.0 (18.3-21.6)
50.4 (48.4-52.5)
9.5 (8.3-10.8)
Abbreviations: ACC/AHA, American College of Cardiology/American Heart Association; ATP-III, Adult Treatment Panel III; ESC, European Society of Cardiology. a
Data are percentage of the population (95% CI) in different categories of treatment recommendations based on the 2013 ACC/AHA,5 2001 ATP-III,4 and 2012 ESC guidelines.6
b
Individuals receiving statin treatment at baseline (n = 315 men and n = 330 women) were excluded.
Women (n = 2315)b
jama.com
JAMA April 9, 2014 Volume 311, Number 14
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
1419
Research Original Investigation
Comparison of Guidelines for CVD Prevention
Figure 2. Observed vs Predicted Risks by the ACC/AHA Risk Model, ATP-III Risk Model, and SCORE Equation Among Rotterdam Study Participants
A ACC/AHA guideline (hard ASCVD)
Men
Women
30
30 Observed risk 25
Predicted risk
20
Hard ASCVD, %
Hard ASCVD, %
25
15 10 5
20 15 10 5
0
0
Original Investigation
Comparison of Application of the ACC/AHA Guidelines, Adult Treatment Panel III Guidelines, and European Society of Cardiology Guidelines for Cardiovascular Disease Prevention in a European Cohort Maryam Kavousi, MD, PhD; Maarten J. G. Leening, MD, MSc; David Nanchen, MD, MSc; Philip Greenland, MD; Ian M. Graham, MD; Ewout W. Steyerberg, PhD; M. Arfan Ikram, MD, PhD; Bruno H. Stricker, MMed, PhD; Albert Hofman, MD, PhD; Oscar H. Franco, MD, PhD Editorial page 1403 IMPORTANCE The 2013 American College of Cardiology/American Heart Association
(ACC/AHA) guidelines introduced a prediction model and lowered the threshold for treatment with statins to a 7.5% 10-year hard atherosclerotic cardiovascular disease (ASCVD) risk. Implications of the new guideline’s threshold and model have not been addressed in non-US populations or compared with previous guidelines. OBJECTIVE To determine population-wide implications of the ACC/AHA, the Adult Treatment Panel III (ATP-III), and the European Society of Cardiology (ESC) guidelines using a cohort of Dutch individuals aged 55 years or older.
Related articles pages 1406 and 1424 Supplemental content at jama.com CME Quiz at jamanetworkcme.com and CME Questions page 1442
DESIGN, SETTING, AND PARTICIPANTS We included 4854 Rotterdam Study participants recruited in 1997-2001. We calculated 10-year risks for “hard” ASCVD events (including fatal and nonfatal coronary heart disease [CHD] and stroke) (ACC/AHA), hard CHD events (fatal and nonfatal myocardial infarction, CHD mortality) (ATP-III), and atherosclerotic CVD mortality (ESC). MAIN OUTCOMES AND MEASURES Events were assessed until January 1, 2012. Per guideline, we calculated proportions of individuals for whom statins would be recommended and determined calibration and discrimination of risk models. RESULTS The mean age was 65.5 (SD, 5.2) years. Statins would be recommended for 96.4% (95% CI, 95.4%-97.1%; n = 1825) of men and 65.8% (95% CI, 63.8%-67.7%; n = 1523) of women by the ACC/AHA, 52.0% (95% CI, 49.8%-54.3%; n = 985) of men and 35.5% (95% CI, 33.5%-37.5%; n = 821) of women by the ATP-III, and 66.1% (95% CI, 64.0%-68.3%; n = 1253) of men and 39.1% (95% CI, 37.1%-41.2%; n = 906) of women by ESC guidelines. With the ACC/AHA model, average predicted risk vs observed cumulative incidence of hard ASCVD events was 21.5% (95% CI, 20.9%-22.1%) vs 12.7% (95% CI, 11.1%-14.5%) for men (192 events) and 11.6% (95% CI, 11.2%-12.0%) vs 7.9% (95% CI, 6.7%-9.2%) for women (151 events). Similar overestimation occurred with the ATP-III model (98 events in men and 62 events in women) and ESC model (50 events in men and 37 events in women). The C statistic was 0.67 (95% CI, 0.63-0.71) in men and 0.68 (95% CI, 0.64-0.73) in women for hard ASCVD (ACC/AHA), 0.67 (95% CI, 0.62-0.72) in men and 0.69 (95% CI, 0.63-0.75) in women for hard CHD (ATP-III), and 0.76 (95% CI, 0.70-0.82) in men and 0.77 (95% CI, 0.71-0.83) in women for CVD mortality (ESC). CONCLUSIONS AND RELEVANCE In this European population aged 55 years or older, proportions of individuals eligible for statins differed substantially among the guidelines. The ACC/AHA guideline would recommend statins for nearly all men and two-thirds of women, proportions exceeding those with the ATP-III or ESC guidelines. All 3 risk models provided poor calibration and moderate to good discrimination. Improving risk predictions and setting appropriate population-wide thresholds are necessary to facilitate better clinical decision making.
JAMA. 2014;311(14):1416-1423. doi:10.1001/jama.2014.2632 Published online March 29, 2014. 1416
Author Affiliations: Author affiliations are listed at the end of this article. Corresponding Author: Maryam Kavousi, MD, PhD, Department of Epidemiology, Erasmus University Medical Center, Rotterdam, PO Box 2040, 3000 CA Rotterdam, the Netherlands (m.kavousi @erasmusmc.nl). jama.com
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
Comparison of Guidelines for CVD Prevention
P
revention of cardiovascular disease (CVD), the leading cause of death worldwide,1,2 remains feasible3 yet suboptimal. The common approach in CVD primary prevention is to identify individuals at high enough risk for cardiovascular events to justify targeting them for more intensive lifestyle interventions, pharmacological interventions, or both. The CVD prevention guidelines developed by the National Cholesterol Education Program expert panel,4 succeeded by the American College of Cardiology/American Heart Association (ACC/AHA) task force5, and the European Society of Cardiology (ESC)6 are the major guidelines influencing clinical practice. While the Adult Treatment Panel III (ATP-III) guidelines were based on the 10-year risk of coronary heart disease (CHD) only,4 the ACC/AHA guidelines broaden to comprise risk of all hard atherosclerotic CVD (ASCVD), including CHD and stroke,5 using the Pooled Cohort equations.7 An additional substantial change in the US guideline is a lower risk threshold for statin treatment in asymptomatic individuals from 20% CHD risk in the ATP-III guidelines4 to 7.5% ASCVD risk in the new guidelines.5 The potential implications of the ACC/AHA guidelines in largely widening the populations endorsed for treatment and the accuracy of the ACC/AHA risk calculator have received much attention.8-12 To be clinically useful, risk prediction models should provide good discrimination. Because decisions for statin treatment are based on an individual’s absolute risk, calibration of the risk prediction models as well as the risk threshold for treatment are important. Varying approaches to CVD risk estimation and application of different criteria for therapeutic recommendations would translate into substantial differences in proportions of individuals qualifying for treatment at a population level. We therefore aimed to determine implications of the ACC/AHA, the ATP-III, and the ESC guidelines in a prospective cohort of Dutch individuals aged 55 years or older. Our first aim was to determine what proportion of the population would be treated based on each guideline. We then sought to examine discrimination and calibration of the 3 risk prediction models underlying these guidelines.
Methods Study Population Analyses were performed within the framework of the Rotterdam Study, a prospective population-based cohort study among persons aged 55 years or older in the Ommoord district of Rotterdam, the Netherlands. The rationale and design of the Rotterdam Study have been described elsewhere.13 The baseline examination took place in 1990-1993 (RS-I). In 2000, the cohort was extended to include inhabitants who reached the age of 55 years in 1990-2000 and persons aged 55 years or older who migrated into the research area (RS-II). The Rotterdam Study was approved by the Medical Ethics Committee of the Erasmus Medical Center and all participants provided written informed consent. The present study used data from the third examination of the original cohort (RS-I, recruited 1997-1999) and the first examination of the extended cohort (RS-II, recruited 2000-
Original Investigation Research
2001). Among the participants aged 75 years or younger, there were 2209 men and 2645 women with measurements required for the analyses. Among these individuals, 315 men and 330 women were receiving statin treatment at baseline and therefore were excluded from the population for whom the eligibility for treatment based on each guideline was assessed. For further analyses on examining the performance of each risk scoring model, exclusions were made using the criteria from each guideline.
Main Outcome Measures and Follow-up Main outcomes were hard ASCVD, composed of fatal and nonfatal myocardial infarction (MI), other CHD mortality, and stroke; hard CHD, composed of fatal and nonfatal MI and CHD mortality; and atherosclerotic CVD mortality.14,15 Prevalent CVD was defined as a history of MI, coronary or other arterial revascularization, stroke or focal transient ischemic attack, or heart failure. Events were assessed until January 1, 2012. A complete description of the methods for measurement of cardiovascular risk factors, definitions of the outcomes, and details regarding the follow-up time is provided in the eAppendix in the Supplement.
Statistical Analyses We calculated the 10-year risk of hard ASCVD events for each individual based on age, systolic blood pressure, treatment of hypertension, total and high-density lipoprotein (HDL) cholesterol levels, current smoking, and history of diabetes mellitus, using the sex-specific parameters from the ACC/AHA Pooled Cohort equations.7 We used the recommended 5% and 7.5% risk thresholds for categorization of the 2 respective categories of “treatment considered” and “treatment recommended.”5 To comply with the ACC/AHA guideline,5 the risk estimation for hard ASCVD was calculated among individuals who were not receiving lipid-lowering medication, were free of CVD at baseline, and had low-density lipoprotein (LDL) cholesterol levels below 190 mg/dL. Using the continuous ATP-III risk prediction model based on age, systolic blood pressure, treatment of hypertension, total and HDL cholesterol levels, and current smoking,16 we also calculated the 10-year risk of hard CHD for the individuals who were not receiving lipid-lowering medication and were free of CVD and diabetes mellitus, to comply with the ATP-III guideline.4 The risk thresholds used for categorization were 10% and 20%, corresponding to the cutoff points for defining the intermediate- and high-risk categories by the ATP-III guideline.4 The 10-year risk of CVD mortality for each participant was based on age, systolic blood pressure, total cholesterol levels, and current smoking using the sex-specific intercepts and regression coefficients from the SCORE equation for low-risk European countries.17 We used the recommended 1%, 5%, and 10% risk thresholds, corresponding to the cutoff points for defining the moderate-risk, high-risk, and very-high-risk groups, respectively, based on the ESC guideline.6,18 To comply with the ESC guideline, the SCORE risk estimation was performed among the individuals who were not receiving lipid-lowering medication at baseline and were free of CVD, diabetes melli-
jama.com
JAMA April 9, 2014 Volume 311, Number 14
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
1417
Research Original Investigation
Comparison of Guidelines for CVD Prevention
Figure 1. Inclusion/Exclusion Criteria for Rotterdam Study Participants for Assessment of Different Guideline Recommendations and Risk Prediction Models ACC/AHA Guideline5
ATP-III Guideline4
4854 Total population 2209 Men 2645 Women
4854 Total population 2209 Men 2645 Women
645 Excluded for statin use at baseline 315 Men 330 Women 4209 Included in treatment recommendations of ACC/AHA guideline 1894 Men 2315 Women
4854 Total population 2209 Men 2645 Women
645 Excluded for statin use at baseline 315 Men 330 Women 4209 Included in treatment recommendations of ATP-III 1894 Men 2315 Women
776 Excluded for prevalent CVD or LDL-C >190 mg/dL (>4.9 mmol/L) 381 Men 395 Women
645 Excluded for statin use at baseline 315 Men 330 Women 4209 Included in treatment recommendations of ESC guideline 1894 Men 2315 Women
802 Excluded for prevalent CVD or DM 463 Men 339 Women
1027 Excluded for prevalent CVD, DM, or CKD 528 Men 499 Women
3433 Included in risk estimation for hard ASCVDa 1513 Men 1920 Women
3407 Included in risk estimation for hard CHDb 1431 Men 1976 Women
3182 Included in risk estimation for CVD mortality 1366 Men 1816 Women
343 Developed hard ASCVD during 10-y follow-up 192 Men 86 Stroke 72 Nonfatal MI 26 Fatal CHD 8 Fatal MI 151 Women 92 Stroke 41 Nonfatal MI 15 Fatal CHD 3 Fatal MI
160 Developed hard CHD during 10-y follow-up 98 Men 65 Nonfatal MI 22 Fatal CHD 11 Fatal MI 62 Women 43 Nonfatal MI 16 Fatal CHD 3 Fatal MI
87 CVD deaths during 10-y follow-up 50 Men 37 Women
ACC/AHA indicates American College of Cardiology/American Heart Association; ASCVD, atherosclerotic cardiovascular disease; ATP-III, Adult Treatment Panel III; CHD, coronary heart disease; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; ESC, European Society of
tus, and chronic kidney disease (CKD).6 Figure 1 describes the inclusion and exclusion criteria for different risk prediction models. Based on each guideline, we formed 3 categories of treatment: “treatment recommended,” “treatment considered,” and “no treatment.” eTables 1 through 3 in the Supplement describe the criteria used to form these 3 treatment categories by each guideline. We assessed the discrimination and calibration of each risk prediction model in our population. Discrimination refers to ability of the model to assign a higher risk to individuals who develop the outcome of interest compared with those who remain free of disease. The discriminative performance of each risk-scoring model was assessed using the C statistic. Calibration is the agreement between the predicted probabilities of disease, based on the risk prediction model, and the actual incidence of events in the population. To assess the calibration of each risk prediction model, the average predicted 10-year risks for each risk function were compared with the average 10-year observed risks (ie, cumulative incidence of the event). Calibration plots were generated to assess the agreement between the predicted and observed risks over the entire range. 1418
ESC Guideline6
Cardiology; LDL-C, low-density lipoprotein cholesterol; and MI, myocardial infarction. a Hard ASCVD includes fatal CHD, nonfatal CHD, and stroke. b Hard CHD includes fatal myocardial infarction, nonfatal MI, and CHD mortality.
Results Baseline characteristics of the participants are presented in Table 1. The mean age of the participants was 65.5 (SD, 5.2) years and 54.5% were women. Based on the ACC/AHA guideline,5 the “treatment recommended” group included 96.4% (95% CI, 95.4%-97.1%; n = 1825) of men and 65.8% (95% CI, 63.8%-67.7%; n = 1523) of women while the “treatment considered” group included 3.3% (95% CI, 2.6%-4.2%; n = 63) of men and 14.2% (95% CI, 12.8%-15.7%; n = 330) of women. Only 0.3% of men (95% CI, 0.1%-0.7%; n = 6) and 20.0% (95% CI, 18.3%-21.6%; n = 462) of women were categorized in the “no treatment” group (Table 2 and eTable 1 in the Supplement). Using the ATP-III guideline, 4 52.0% (95% CI, 49.8%54.3%; n = 985) of men and 35.5% (95% CI, 33.5%-37.5%; n = 821) of women were categorized in the “treatment recommended” group, while the “treatment considered” group included 14.2% (95% CI, 12.6%-15.8%; n = 269) of men and 14.1% (95% CI, 12.7%-15.6%; n = 326) of women. The “no treatment” category included the remaining 33.8% (95% CI,
JAMA April 9, 2014 Volume 311, Number 14
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
jama.com
Comparison of Guidelines for CVD Prevention
Original Investigation Research
31.7%-35.9%; n = 640) of men and 50.4% (95% CI, 48.4%52.5%; n = 1168) of women (Table 2 and eTable 2 in the Supplement). Based on the ESC guideline, 6 66.1% (95% CI, 64.0%68.3%; n = 1253) of men and 39.1% (95% CI, 37.1%-41.2%; n = 906) of women were included in the “treatment recommended” category. The “treatment considered” group comprised 31.6% (95% CI, 29.5%-33.7%; n = 598) of men and 51.4% (95% CI, 49.3%-53.4%; n = 1189) of women. Only 2.3% (95% CI, 1.6%-2.9%; n = 43) of men and 9.5% (95% CI, 8.3%-10.8%; n = 220) of women were assigned to the “no treatment” category (Table 2 and eTable 3 in the Supplement). eFigure 1 in the Supplement presents the treatment recommendations based on the 3 guidelines for the populations younger than 65 years and aged 65 years or older. The data suggest that almost all men older than 55 years and nearly all women older than 65 years are recommended for statin treatment based on the new ACC/AHA guideline. eTables 1 through 3 in the Supplement show that while all men and women with prevalent CVD were categorized in the “treatment recommended” group by the ACC/AHA guideline (eTable 1 in the Supplement), 12.9% of men and 4.2% of women with clinical CHD and CHD risk equivalents were categorized in the “treatment considered” or “no treatment” category based on the ATP-III guideline (eTable 2 in the Supplement). Using the ESC guideline, a small group of individuals with clinical CVD and its risk equivalents (0.6% of men and 0.4% of women) were categorized in the “treatment considered” group (eTable 3 in the Supplement). eTables 4 through 6 in the Supplement provide the description of the proportion of the population to whom each risk estimation model was applied. Among 1513 men and 1920 women included for ASCVD risk prediction (ACC/AHA), 192 men and 151 women developed hard ASCVD over 10-year follow-up. Among 1431 men and 1976 women included for CHD risk prediction (ATP-III), hard CHD occurred in 98 men and 62 women over 10-year follow-up. Among 1366 men and 1816 women included for CVD mortality risk prediction (ESC), 50 men and 37 women died of atherosclerotic CVD over 10-year follow-up. For all outcomes studied, follow-up time was truncated at 10 years for individuals with a longer follow-up time than 10 years. After calculating the 10-year risk for individuals based on each risk prediction model, we first assessed the discriminative ability of each model. The C statistic for the ACC/AHA
model was 0.67 (95% CI, 0.63-0.71) for men and 0.68 (95% CI, 0.64-0.73) for women for hard ASCVD. Use of the ATP-III risk prediction model resulted in a C statistic of 0.67 (95% CI, 0.620.72) for men and 0.69 (95% CI, 0.63-0.75) for women for hard CHD. Using the SCORE equation (ESC), the C statistic was 0.76 (95% CI, 0.70-0.82) for men and 0.77 (95% CI, 0.71-0.83) for women for CVD mortality. We then assessed the calibration of each risk prediction model. Figure 2 compares the average 10-year risks predicted by the ACC/AHA, ATP-III, or SCORE (ESC) risk prediction models with the observed 10-year risks (ie, cumulative incidence of events) in each risk category. Calibration was poor for all 3 models; the ACC/AHA (Figure 2A), the ATP-III (Figure 2B), and the SCORE equation (Figure 2C) overestimated the 10-year risk among men and women across all risk categories. eTable 7 in the Supplement details the percentage of population at different categories of risk using each risk prediction model. The average predicted risks vs observed cumulative incidence of
Table 1. Characteristics of the Study Population at Baseline Men (n = 2209)
Characteristics Age, mean (SD), y
65.5 (5.3)
65.4 (5.2)
Systolic
143 (21)
140 (21)
Diastolic
79 (11)
76 (11)
468 (21.2)
643 (24.3)
Blood pressure, mean (SD), mm Hg
Antihypertensive treatment, No. (%) Body mass index, mean (SD)a
26.7 (3.3)
27.3 (4.5)
Total cholesterol, mean (SD), mg/dL [mmol/L]
216.2 (37.1) [5.60 (0.96)]
232.7 (35.7) [6.03 (0.92)]
HDL cholesterol, mean (SD), mg/dL [mmol/L]
47.7 (12.1) [1.24 (0.31)]
58.1 (14.9) [1.50 (0.39)]
LDL cholesterol, mean (SD), mg/dL [mmol/L]
140.2 (34.4) [3.63 (0.89)]
147.9 (34.4) [3.83 (0.89)]
Statin treatment at baseline, No. (%)b
315 (14.3)
330 (12.5)
Current smoking, No. (%)
437 (19.8)
522 (19.7)
Diabetes mellitus, No. (%)
315 (14.3)
282 (10.7)
Chronic kidney disease, No. (%)
139 (6.3)
226 (8.5)
Prevalent CVD, No. (%)
414 (18.7)
186 (7.0)
Abbreviations: CVD, cardiovascular disease; HDL, high-density lipoprotein; LDL, low-density lipoprotein. a
Calculated as weight in kilograms divided by height in meters squared
b
No. (%) of men and women receiving lipid-lowering medication at baseline. (Statins constituted 96% of all lipid-lowering medications at baseline).
Table 2. Treatment Recommendations Based on Different Guidelines Guidelinea Treatment Categories
ACC/AHA
5
ATP-III4
ESC6
52.0 (49.8-54.3)
66.1 (64.0-68.3)
b
Men (n = 1894)
Treatment recommended
96.4 (95.4-97.1)
Women (n = 2645)
Treatment considered
3.3 (2.6-4.2)
14.2 (12.6-15.8)
31.6 (29.5-33.7)
No treatment
0.3 (0.1-0.7)
33.8 (31.7-35.9)
2.3 (1.6-2.9)
Treatment recommended
65.8 (63.8-67.7)
35.5 (33.5-37.5)
39.1 (37.1-41.2)
Treatment considered
14.2 (12.8-15.7)
14.1 (12.7-15.6)
51.4 (49.3-53.4)
No treatment
20.0 (18.3-21.6)
50.4 (48.4-52.5)
9.5 (8.3-10.8)
Abbreviations: ACC/AHA, American College of Cardiology/American Heart Association; ATP-III, Adult Treatment Panel III; ESC, European Society of Cardiology. a
Data are percentage of the population (95% CI) in different categories of treatment recommendations based on the 2013 ACC/AHA,5 2001 ATP-III,4 and 2012 ESC guidelines.6
b
Individuals receiving statin treatment at baseline (n = 315 men and n = 330 women) were excluded.
Women (n = 2315)b
jama.com
JAMA April 9, 2014 Volume 311, Number 14
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Purdue University User on 05/28/2015
1419
Research Original Investigation
Comparison of Guidelines for CVD Prevention
Figure 2. Observed vs Predicted Risks by the ACC/AHA Risk Model, ATP-III Risk Model, and SCORE Equation Among Rotterdam Study Participants
A ACC/AHA guideline (hard ASCVD)
Men
Women
30
30 Observed risk 25
Predicted risk
20
Hard ASCVD, %
Hard ASCVD, %
25
15 10 5
20 15 10 5
0
0
